Henning Kolbjørnsen,
National Metrology Laboratory
Justervesenet (Norwegian Metrology Service)
Evaluation of measurement uncertainty
in transferred LNG volume to or from a
ship tank
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Focus in this presentation
Report:
Evaluation uncertainty in transferred LNG volume
VSL (P. Lucas), FORCE (L. Poder), JV (T. Mortensen, K. Hellerud, H. Kolbjørnsen)
(Delivered ultimo 2011)
• The need for a common metrological language and understanding and
methodology for uncertainty assessment.
• Some reflections to the different types of tanks, level gauging equipment,
trim/list measurements etc.
• The users need for understanding their measurement system and its use.
• Traceability in metrology for custody transfer
• Some significant differences between applied systems
• Visualization of significant contributions to the combined uncertainty
• Propagation of uncertainty trough correct models (model functions)
• Uncertainty in estimated unloaded volume from a complete vessel
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Tank level gauging of LNG –
a challenging metrological task
The measurements should be transparent and have a high degree of confidence
for all parties involved. The measurement result should be reported with a
measurement value and the associated expanded measurement uncertainty.
(Un)Loaded volume of LNG = vLNG ± Uv,LNG
The measurement result should be reported in a way that give trust that all
relevant input quantities are identified and included in the result reported.
Important:
• This result should clearly be based on the actual measuring method
and system in use.
• The results should include all relevant influences and operational
effects and be corrected for known systematic errors.
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Metrology terms according to the
international vocabulary on metrology OK (VIM available at www.BIPM.ORG)
Metrological traceability (VIM 2.43):
- Obtained through an unbroken chain of
calibrations to a common reference system for all
measurements.
- The references are normally maintained by
national metrology institutes or BIPM
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Metrology terms according to the
international vocabulary on metrology OK (VIM available at www.BIPM.ORG)
Calibration (VIM 2.39):
A calibration is the comparison of an instrument with a metrological reference
to establish the value of the instrument. In this way a common reference
system is secured.
Note:
• Adjustment should not be confused with calibration. The confusing term
“self-calibration” should never be used as it violate the understanding of
the need for traceable measurements
• Calibrations will normally be repeated to obtain knowledge of the stability
of the instrument in use. Information on stability may alternatively be
obtained based on other identical instruments use under the same
conditions.
• Measuring instruments should always be maintained in according to
planned interval based on knowledge of the instruments stability.
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Some relevant input quantities for tank
based measurements
An overview of relevant input quantities is given. Red color indicates that they may
significantly influence the measurement of transferred volume.
Tank
Calibration
Drift/stability
Resolution
Temp. dim.
structure
Hydrostatic
pressure
Sagging/
Hogging
Inclinometer
Calibration
Drift/stability
Resolution
Disper. in
readings
Sagging/
Hogging
Pressure gauge
Calibration Drift
Float level gauge
Tape temp.
Liquid density
Boyancy
Calibration Location
Drift
Disper. in
readings
Temp.
sensors
Calibration
Drift/stability
Disper. in
readings
Radar level gauge
Calibration
Mount. position
Temp.
Drift
Disper.
readings
Surface detection
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Stability in tank dimensions –
what do we know?
Input quantities related to possible changes in size and shape of
tanks should have a high focus related to potential impact:
• Recalibration of ship tanks only done after repair / changes of dead volume.
The work of NBS (LNG Measurements, NBSIR 85-3028) seems to be the most
significant work done on the stability of ship tanks of different types.
Available validations will increase transparency and document uncertainty
estimates.
• Results of relevant inter-comparisons between calibration suppliers for
gauge tables will demonstrate a common reference level and give increased
confidence in gauge tables.
• Thermal deformation, loading pressure and sagging/hogging will
change the dimensions of the tank dependant on type of tank.
Ref: www.tu.no
Ref: www.worldwideflood.com
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Temperature effects on tank volume
Temperature effects on the ship tanks should be addressed according to their
construction and material used.
Membrane tanks
• Thermal barrier between the LNG/NG and supporting structure.
• Supporting structure expected to stabilize at a temperature
close to the temperature of ballast water.
• Gauge tables for membrane tanks use 20 C as ref.
temperature
Moss tanks
• Tanks are self support around equator.
• Isolated is outside the dimensional structure so that the
change in the volume relates to the int. temperature
(liquid/vapor).
• The reference temperature for the gauge table is -160 C.
Ref: www.twi.co.uk
Ref: www.tu.no
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Level gauging
A multitude of influences related to level gauging should be identified and accounted
for:
• Level gauges will measure height with reference to top of tank. Thermal or
mechanical changes may change the height of tank affecting this reference level.
• Thermal effects on level gauge from interior temperature (liquid/vapor) must be
corrected for. Errors from these effect will be correlated to errors related to the
thermal changes of top reference level on tank.
• Calibration methods for different level gauges must be identified, and they should
be made according to expected drift of instrument.
• Primary and secondary level gauge should not serve as reference for each other.
• Significant deviations of reading of primary and secondary level gauge should
abandon measurements
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Sensitivity of errors in level measurements and
trim/list as a function of level
The inclinometer may in some situations be critical for the correct measurement of
loaded / unloaded volume.
• The sensor should be mounted / corrected to the horizontal plane defined for the
tank gauge table. It should be emphasized that it should indicate the trim and list
for the specific tank being filled or emptied.
• Sagging/hogging may introduce errors between reference plane for tank and
mounting position of inclinometer.
• Calibration, resolution and repeatability of readings may be relevant to the
uncertainty analysis.
• The sensitivity of an error in the inclinometer
reading will significantly be influence by the
operating situation .
Ref: www.directindustry.com
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Uncertainty assessment
Some important precondition for the uncertainty assessment:
• The propagation of uncertainty is only possible if there is a defined model
function or the sensitivity in the result (output quantity) for a given input
quantity is given. The model will normally depend on the method.
• The uncertainty calculation is done related to one specific operating state.
Relative uncertainties can only be combined directly if the model function is
built up by multiplications/divisions. This is not the case for model functions
related to tank gauging of LNG.
• The input quantities and their uncertainty estimates should be related to the
operation conditions, the actual instruments used and operational procedures.
• Significant correlations between input quantities should be accounted for.
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
A measurement model is needed
For each tank gauging system a mathematical models must be identified. The model
should be used for calculating the volume given estimates for all “input quantities”
• Readings (often the mean of repeated readings)
• Corrections from last calibrations and drift
• Corrections due to thermal, mechanical or constructional effects
• Requirements for maximum list and trim during loading and stabilizing time
• Corrections for non reference conditions like for the temperature in the supporting
structure of a membrane tank
• …
Detailed model functions depend on each type of system. They are not given in the
“Custody Tranfere Handbook” given by GIIGNL. Below som functions used for
propagating uncertaint in the project :
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Sensitivity coefficients in the measurement
method must be known
The sensitivity coefficients are given by the model function and the estimates for
the input quantities.
Below are some examples on how sensitive coefficients relates
to different operating situations like:
• Both for membrane and Moss type tanks, the liquid surface area of the liquid
will decrease for low and high levels of LNG. This is especially the case for a
spherical shaped tank. Partially loading/unloading will as a consequence have
a higher combined uncertainty.
• Influences of deviations from “even keel” (zero trim and list) in level
measurements caused by a horizontally location not in the center of the tank
or with an angel to the horizontal plane, will be more significant as trim and list
deviates from zero.
Ref:ntnu.no
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Correlation of input quantities
Input quantities may be correlated in some degree. This will influence the combined
uncertainty. Positive correlation in input quantities which are added will increase the
combined uncertainty.
JCGM 100:2008 (GUM 1995 with minor corrections)
Evaluation of measurement data — Guide to the expression of uncertainty in
measurement
Paragraph 5.2.4 states that there may be significant correlation between two input
quantities if the same measuring instrument, physical measurement standard, having a
significant standard uncertainty is used in their determination.
Paragraph 5.2.5 states that correlations between input quantities should not be ignored if
present and significant.
It may be difficult to evaluate the degree of correlation. Taking correlation into
account in a conservative way is preferred.
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Presentation of an uncertainty analysis
Given a model function an uncertainty analysis may be presented in a
format like the below. This can be efficient for identifying and improving
the measurement system / method.
Input quantety /
parameter Value
standard
uncertainty
relative standard
uncertainty sensitivity contribution
Vtable (Trunc(hstart)) 34000 68 0.20% 1 68.0
hstart 22.90 0.015 0.07% -1273 -19.2
DVSaggingHogging,start 0 34 NA 1 34.0
DVHydrostatic,start 0 34 NA 1 34.0
DVTable,drift,start 0 8.5 NA 1 8.5
Vtable (Trunc(hstop)) 1600 3.2 0.20% 1 3.2
hstop 0.15 0.0063 4.19% -19 -0.1
DVSaggingHogging,stop 0 1.6 NA 1 1.6
DVHydrostatic,stop 0 1.6 NA 1 1.6
DVTable,drift,stop 0 0.40 NA 1 0.4
Ttank,start (°C) 20 10 50% 1 11.2
Ttank.stop (°C) 20 10 50% 0 0.5
Ttank,ref (°C) 20 1.0 5.0% -1 -1.1
a 1.10E-05 0.00 5.0% 0 0.0
Vtank unloaded 33242 uVloaded 87.97
UV,loaded 175.95
UV,loaded* 0.53% (Note: Values given as examples to clarify method)
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Maintenance of each measurement system
Every measurement system needs skilled operators, following up and maintenance:
• There should be a clear understanding of the responsibility and needed
competence for the LNG measurement system at each ship.
• An updated version of the uncertainty evaluation given in a easy readable format
should be available for each loading/unloading. From this it should be possible to
identify the most significant contributions to the actual measurement.
• The uncertainty budget and its estimates should be clearly linked to the actual
system in use and the conditions during the measurement.
• The measurement result included the measurement uncertainty (95 % confidence)
should be reported.
• Based on experience with each system it should be possible to improve
measurements and reduce measurement uncertainty
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Uncertainty in unloading a vessel
For unloading a vessel, the unloading volume of each tank is summed to a total:
• Same calibration reference, methods etc.
• Same operational /maintenance procedures
• Same environmental situation
Model function divided into two input summarized quantities for each tank. One
summarize correlated input quantities and one uncorrelated input quantities.
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Uncertainty in unloading a vessel (cont,)
High degree of correlation for measurement of (un)loaded volume from/to a vessel
(Note: Values given as examples to clarify method)
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Summary
Model functions reflecting the actual measurement method (tank, level gauge, relevant
temperature measurements, trim/list measurement and errors related to
traceability/calibrations should be maintained for each ship. The most significant
contributions to the measurement uncertainty should be identified and if possible
reduced.
Temperature effects caused by deviations from reference conditions for tank (interior
or exterior) on tank volume, height of reference point for level gauge and the level gauge
itself should be carefully accounted for with corrections and uncertainty estimates.
Errors in trim and list may significantly influence volume measurements dependant on
operational conditions, type of tank and mounting position of sensor. Both errors and
observed fluctuations should be accounted for in the uncertainty estimate. Care related
to the mounting position, maintenance and calibration of the sensors for measurement
of trim and list is important. Preferably it should be measured locally for each tank and
synchronized with the level measurement.
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Summary (cont.)
Correlation effects between input quantities, for instance related to corrections for
temperature effects, must be accounted for. They may significantly increase or
decrease the combined measurement uncertainty.
Data from validation or other investigations, can increase understanding of tank
gauging of LNG volume significantly and should be shared to increase knowledge. It is
a need for obtaining more information on stability of instruments/tank and influences.
Comparable measurements systems will increase knowledge. Measurement
comparisons made a
An updated version of the uncertainty evaluation given in a easy readable format
should be available for each loading/unloading. The person responsible for the
measurement system should use this tool to identify the most significant contributions
to the actual measurement and if needed initiate improvements.
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Summary (cont.)
Based on experience with each system it should be possible to improve
measurements and reduce measurement uncertainty
Improvements in the calibration method for LNG tanks will contribute directly to
reduced uncertainty. Intercomparissons should be performed to document these
improvements.
More documentation of the stability of both membrane tanks and self supported
spherical tanks would be preferable.
Identified known and systematic effects should be corrected for to reduce the
uncertainty. This may relate to temperature effect and drift in measurement equipment.
Metrology for LNG, Final Conference, 17. &18.. October 2013, Delft, the Netherlands
Thank you for your attention!